Hello, I have a test rh el 6 server where I have 6 network adapters, but I think my idea could also cover for example an F13 server with virt-preview repo... eth0 and eth1 are dedicated for bond0 and main ip for kvm host eth4 and eth5 are dedicated as multipath iSCSI initiators for configuring storage pools on disks presented by an iSCSI SAN array
I would like to dedicate eth2 and eth3 to network for my guests, but I want to cover multiple vlans. Suppose I want to cover vlanid=1 and vlanid=2 Here is my planned approach 1) configure bond1 with eth1 and eth2 as slaves Something like # cat ifcfg-eth2 DEVICE="eth2" HWADDR="xxxxxxxxxxxxx" NM_CONTROLLED="no" ONBOOT="yes" SLAVE=yes MASTER=bond1 MTU=9000 --> is it ok to put MTU line here or do I have to put it at bond interface definition level? 2) bond1 is without ip/netmask information; something like DEVICE=bond1 BOOTPROTO=none ONBOOT=yes TYPE=Ethernet 3) create bond1.1 script something like this: DEVICE=bond1.1 ONBOOT=yes BOOTPROTO=none VID=1 VLAN=yes IPADDR=myip.1 NETMASK=mynet.1 BROADCAST=mybroadcast.1 BRIDGE=brvlan1 4) create bond1.2 script something like this: DEVICE=bond1.2 ONBOOT=yes BOOTPROTO=none VID=2 VLAN=yes IPADDR=myip.2 NETMASK=mynet.2 BROADCAST=mybroadcast.2 BRIDGE=brvlan2 5) create brvlan1 script DEVICE=brvlan1 TYPE=Bridge BOOTPROTO=dhcp ONBOOT=yes DELAY=0 6) create brvlan2 script DEVICE=brvlan2 TYPE=Bridge BOOTPROTO=dhcp ONBOOT=yes DELAY=0 QUESTION: is it necessary to put BOOTPROTO=dhcp for bridge interfaces? Can I put a static ip or better no ip at all? 7) service network restart with eventually adjustments fro iptables 8) service libvirtd reload If the steps above make sense 9) If I want a guest on vlan1 I have to assign it shared physical device --> brvlan1 QUESTION: in guest do I have to tag the lan or should it be transparent with the guest? Otherwise: waht would be a correct different way of proceeding? If my approach is correct, would it make sense to have similar approache for bond0, so that I can use this bond interface for guests too? Thanks in advance. Gianluca _______________________________________________ virt mailing list [email protected] https://admin.fedoraproject.org/mailman/listinfo/virt
